Biochemical changes of the pericellular matrix and spatial chondrocyte organization—Two highly interconnected hallmarks of osteoarthritis

During osteoarthritis, chondrocytes change their spatial arrangement from single to double strings, then to small and big clusters. This change in pattern has recently been established as an image‐based biomarker for osteoarthritis. The pericellular matrix (PCM) appears to degrade together alongside cellular reorganization. The aim of this study was to characterize this PCM‐degradation based on different cellular patterns. We additionally wanted to identify the earliest time point of PCM‐breakdown in this physiopathological model. To this end, cartilage samples were selected according to their predominant cellular pattern. Qualitative analysis of PCM degradation was performed immunohistochemically by analysing five main PCM components: collagen type VI, perlecan, collagen type III, biglycan, and fibrillin‐1 (n = 6 patients). Their protein content was quantified by enzyme‐linked immunosorbent assay (127 patients). Accompanying spatial cellular rearrangement, the PCM is progressively destroyed, with a pericellular signal loss in fluorescence microscopy for collagen type VI, perlecan, and biglycan. This loss in protein signal is accompanied by a reduction in total protein content from single strings to big clusters (P < .001 for collagen type VI, P = .003 for perlecan, and P < .001 for biglycan). As a result of an increase in the number of cells from single strings to big clusters, the amount of protein available per cell also decreases for collagen type III and fibrillin‐1, where total protein levels remain constant. Biochemical changes of the PCM and cellular rearrangement are thus highly interconnected hallmarks of osteoarthritis. Interestingly, the earliest point in time for a relevant PCM impairment appears to be at the transition to small clusters.


| INTRODUCTION
Osteoarthritis (OA) is a common degenerative joint disease characterized by the irreversible destruction of the articular cartilage that affects a big proportion of the elderly population. Since there is no defined degenerative cut-off threshold to formulate the diagnosis of OA, the condition is difficult to quantify. It has recently been shown that the chondrocytes within the articular cartilage are arranged in distinct spatial patterns along the arciform collagen fiber network, 1 following a vertical orientation in the ascending/ descending part of the collagen arcades, and a horizontal orientation in the apex of the arcs in the superficial zone. 2 The predominant physiological cellular pattern found in the human femoral condyle is that of single strings. 1 With the onset of OA, the cellular spatial organization changes by rearranging first from single to double strings, 3 followed by small and finally by big clusters. 4 Interestingly, these spatial changes are not only a hallmark of OA, but they also seem to depend on, and to be highly specific to localized tissue degeneration. Big clusters, in particular, are, for example, often localized near cartilage fissures in the upper layer. 4 Spatial cellular organization has therefore been suggested for use as an image-based biomarker for local tissue degeneration at the cellular level. 5 In the immediate surrounding of the chondrocytes, a unique and highly specialized area is present, termed the pericellular matrix (PCM). It has been suggested that this area acts as a mechanosensitive cell-matrix interface, 6 protecting the chondrocytes from apoptosis and modulating the cellular biosynthetic response to outside stimuli. 7 Interestingly, the PCM is not particular and exclusive to articular cartilage, but it seems to be a general feature of connective tissue: the presence of a PCM has also been described in the intervertebral dis, 8 temporomandibular cartilage and disc, 9 meniscus, 10 tendons, 11 and cricoarytenoid cartilage. 12 Perlecan (also known as heparan sulfate proteoglycan 2), 13 collagen VI, 14 biglycan, 15 fibrillin-1, 16 collagen type IX, 17 and collagen type III 18 are key components of the PCM.
With OA onset and progression, a cascade of inflammatory and catabolic processes is triggered, which leads to an upregulation as well as elevated activity of various matrix metalloproteinases (MMPs). 19 This ultimately triggers collagen fibril denaturation 20 and loss of proteoglycans. 21 Of note, even though an upregulation in the collagen expression 22 is present in OA-affected cartilage, an overall decrease of collagen type II of the extracellular matrix is observed, particularly in the upper fibrillated area of the advanced osteoarthritic samples. 23 Similar observations have been made for PCM-collagen type VI. 24 It can thus be assumed that the PCM is destroyed during the course of OA, 24 which has also been suggested for the PCM in the intervertebral disk 25 and meniscus. 26 Collagen type VI and perlecan are thus well established and investigated PCM components in both healthy and OA affected tissue 24,27 ; several studies have emphasized their involvement and connection to the biomechanical properties of the PCM. 28,29 Little information is currently available with respect to other minor compositional PCM constituents, such as collagen type III, biglycan, and fibrillin-1.
Collagen type III has been shown to be present in a diffuse pattern in the pericellular region, in the close vicinity of the chondrocytes within the articular cartilage, as well as in the intervertebral disc. 30 Even though collagen type III is a feature of tissue repair in damaged tissue, low amounts of collagen type III are detected in both normal adult cartilage and OA-affected cartilage. 30,31 Similar observations have been made for biglycan, which has been described to have a pericellular localization 32 where it is directly attached to the collagen fibrils. 32 This may indicate its possible involvement in modulating morphogenesis and differentiation. 33 Also, pericellularly localized, fibrillin-1 regulates the viscoelastic properties of the tissue, 34 as it is involved in the cell-matrix mechanotransduction 35 and the cell-matrix adhesion. 36 Fibrillin-1 has been shown to interact with perlecan in weight-bearing tissues by regulating the bioavailability of transforming growth factor-β1 37 and playing a role in bone morphogenetic protein-signalling. 38 In order to paint the big picture with respect to PCM degradation and to better understand the underlying mechanisms, these minor and often overlooked PCM components and the synergetic role they play in the degradative process of the pericellular region need to be elucidated. In a previous study, we had already shown that the PCM progressively loses its characteristic biomechanic properties alongside cellular rearrangement. 39 We had further demonstrated that, accompanying these pattern changes, the two main PCM components-collagen type VI and perlecan-undergo progressive destruction. 39 In the present study, the current knowledge of PCM degradation on the basis of cellular spatial organization is further expanded upon by additionally investigating its components collagen type III, biglycan, and fibrillin-1. We also wanted to identify the point in time in the physiopathological model of spatial cellular rearrangement when the PCM structure breaks down. The hypothesis was that the PCM integrity is progressively disrupted during the course of cellular rearrangement, and that total protein content decreases with increasingly pathological spatial cellular arrangement. We also analyzed whether or not the observed breakdown of the PCM is a generalized process occurring within the entire tissue or only within isolated superficial areas of the cartilage.  inflammatory pathology (eg, rheumatoid arthritis) were completely excluded. Samples were collected from a total of 139 patients, with 78 women (aged 67 (35-88) years) and 61 men (68  years) (for sex-age distribution and frequency, see Figure S1). Articular cartilage coming exclusively from femoral condyles was used for the immunohistological and biochemical analyses in the present study ( Figure S2A).

| Cartilage immunostaining
Cartilage sections for both side views and top-down views were processed and labeled for collagen type VI, perlecan, collagen type III, biglycan (n = 6 patients with paraformaldehyde-fixed cartilage), and fibrillin-1 (n = 6 patients unfixed cartilage). As fibrillin-1 is a calcium-binding protein, 40 the ethylenediaminetetraacetic acid decalcification procedure used for the side-view analysis led to an overall loss of immunoreactivity. No decalcification process was therefore used for fibrillin-1 staining; rather, the cartilage was harvested from the bone directly. Histological sections were pretreated with 0.2% (w/v) collagenase type XI (Sigma-Aldrich, Taufkirchen, Germany) for collagen type VI and fibrillin-1 staining and with 0.1% (w/v) hyaluronidase (Sigma-Aldrich) for perlecan, collagen type III, and biglycan staining in PBS for 1 hour at 37°C, followed by three washing steps with PBS. To reduce unspecific antibody binding, sections were blocked for 30 minutes in 5% (w/v) bovine serum albumin and 0.3% (v/v) Triton X-100 in PBS. This was followed by incubation with the primary monoclonal antibodies at a dilution of 1:100 in 2.5% (w/v) bovine serum albumin-PBS at 4°C overnight: collagen type VI (rabbit anticollagen VI, ab-182744; Abcam, Cambridge, UK), perlecan (mouse anti-perlecan, sc-377219; Santa Cruz Biotechnology Inc., Dallas, TX), collagen type III (rabbit anti-collagen type III, ab-7778;

| Optical assessment of PCM integrity
The analysis for structural integrity of the PCM was performed using a blinded optical approach to the histological side views, as previously described by Hofmann et al. 25 Due to the arciform collagen fiber and cell orientation, these sections allow for better visualization of the developing patterns, especially in deeper cartilage layers where the orientation of the cells is perpendicular to the articular surface rather than parallel to it as in the superficial zone.
Two rectangular regions of interest of approximately 500 × 500 μm were selected at random from each mosaic image per cellular pattern.
PCM integrity assessment was performed as previously described. 25 The percentage of the circumference of the cell that was surrounded by the staining signal for the PCM was thereby determined for the immunolabelled PCM components collagen type VI, perlecan, collagen type III, biglycan, and fibrillin-1. The percentages of the PCMs that were encountered were ordinally classified into one of the following three groups: less than 25%, 25% to 75%, and more than 75%. A weighted arithmetic mean was then calculated for each staining and section for further analysis. To quantify protein content of the specific PCM components, sandwich ELISAs were performed for each distinct cellular pattern.
For collagen type III, a total of 10 µg of protein was used. For biglycan (BIOZOL Diagnostica Vertrieb GmbH, Eching, Germany) and fibrillin-1 (LifeSpan Biosciences Inc., Seattle), 30 µg was used following the manufacturer's protocol. For collagen type VI and perlecan, the data from a previous publication were used. 39

| Statistical analysis
The values obtained from the two regions of interest per pattern and side-view histological section were averaged for further analysis.

| RESULTS
In our top-down histological qualitative analysis, in healthy tissue areas ascribed to single strings, the PCM signal was intact and welldefined for collagen type VI, perlecan, collagen type III, biglycan, and fibrillin-1 staining ( Figure 1B  rather it was scattered randomly between the cells. Interestingly, signal intensity remained constant for collagen type III and fibrillin-1, although the structural integrity of the PCM also appeared to be compromised ( Figure 1D 1 -D 4 ,F 1 -F 4 ).
To further elaborate these findings, the PCM signal was analyzed in side views of the cartilage (Figures 2 and 3). While no major changes could be observed from single to double strings, the number of cells per tissue area increased slightly from double strings to small clusters, and increased greatly from small to big clusters (Figure 3). This increase in the number of cells was accompanied by a reduction in the number of cells that were encompassed by an intact PCM. At the same time, the number of cells with a scattered PCM increased.
Moreover, when analyzing the averaged PCM integrity per cell based on their cellular organization, a significant decrease of collagen type VI, biglycan and fibrillin-1 ( Figures 4A and 4D,E, Table 1) was observed. Perlecan and collagen type III signaling around the cells also decreased, without, however, reaching statistical significance ( Figure 4B,C).
To measure the total protein content in the tissue, ELISAs were performed. Complementary to the results from the immunohistochemical analysis, the quantitative ELISAs showed that the more pathological the cellular pattern, the more the protein content in collagen type VI, perlecan, and biglycan decreased ( Figures 5A, 5B, and 5D). The earliest point in the sequence of events at which a significant reduction could be observed was at the transition from single strings to small clusters for collagen type VI (P = .016) and perlecan (P = .008), and at the transition F I G U R E 2 Evaluation of signal presence in the pericellular matrix (PCM) and optical readout in immunohistochemical sections. Side view of the cartilage with its underlying subchondral bone after perpendicular sectioning and immunolabelling. Representative areas were selected for optical readout of PCM integrity after PCM staining (in this case, collagen type III antibody) (white) and nuclear staining with 4′,6-diamidino  Table 2). While the lowest amounts of protein were measured in big clusters for all of the aforementioned PCM components, collagen type III and fibrillin-1 displayed a constant total protein content throughout all subsequent cellular rearrangements ( Figures 4C and 4E).

| DISCUSSION
The aim of the present study was to investigate the degeneration of the PCM on the basis of local cellular spatial organization by analysing five main components of the pericellular region: collagen type VI, perlecan, collagen type III, biglycan, and fibrillin-1. We hypothesized that the PCM becomes progressively structurally impaired and that the protein content of these components is concomitantly reduced. The progressive impairment of collagen type VI and perlecan with increasingly pathological spatial organization had been described previously, 24,39 and thus served as a reference experiment. It was shown previously that collagen type VI, which is initially pericellularly well defined in strings and in double strings, shows a much more diffuse signal in big clusters, where its intracellular signal is lost. Perlecan in strings has a signal both extracellularly and intracellularly. Both signals are mostly lost in degenerative cartilage. The staining signals of the antibodies for collagen type III and fibrillin-1, however, are only slightly impaired in their structural integrity, and they remain at a relevant signal intensity. Biglycan is initially present in the PCM and has a slight intracellular signal. Later, its fluorescence signal is no longer present immediately surrounding the cells, and its total protein content is reduced. The signal now encircles the lacuna in big clusters at the interface to the interterritorial matrix. Moreover, the biglycan signal directly adjacent to the nucleus greatly increases. Since the PCM is evidently disrupted in the course of OA, the question arises: why are its collagen type VI and perlecan components so clearly being reduced, while collagen Since proteolytic enzymes are elevated in osteoarthritic cartilage, 19 a loss of structural proteins such as collagen type VI and perlecan only appears natural and can be considered to be the synergetic result of the various catabolic MMPs. The loss of perlecan F I G U R E 4 Changes in average pericellular matrix (PCM) integrity are associated with cellular spatial rearrangement. Boxplots showing the calculated weighted arithmetic means for the PCM for collagen type VI (A), perlecan (B), collagen type III (C), biglycan (D), and fibrillin-1 (E) for each of the cellular organizational patterns. A significant decrease for collagen type VI, biglycan, and fibrillin-1 was observed with cellular reorganization. For perlecan and collagen type III, a decreased tendency was observed but without reaching statistical significance. *P < .05. BC, big clusters; DS, double strings; SC, small clusters; SS, single strings Mann-Whitney U test. 2176 | F I G U R E 5 Quantification (ELISA) of the five major pericellular matrix components (collagen type VI, perlecan, collagen type III, biglycan, and fibrillin-1) analyzed according to the locally dominant cellular spatial organization. Bar diagrams displaying protein content as measured by ELISAs with homogenized cartilage for collagen type VI (A), perlecan (B), collagen type III (C), biglycan (D), and fibrillin-1 (E). Prior to analysis, cartilage samples had been grouped according to the locally predominant spatial cellular pattern. From single strings to big clusters, a significant decrease in collagen type VI, perlecan, and biglycan can be observed.
No changes with spatial cellular rearrangement were observed in the content of collagen type III and fibrillin-1. Data displayed as mean ± standard deviation. Images (A) and (B) taken from. 39 *P < .05; **P < .01; ***P < .  39 This is still a stage where the cartilage appears to be macroscopically intact.
If it is assumed that an intact functional unit is the prerequisite for intact cartilage metabolism and resilience, and that in bigger

| STUDY LIMITATIONS
The results of the mathematical analysis of the PCM derived from the histological sections presented in this study are based on a highly subjective procedure. We, therefore, limited the variable to an ordinal scale of only three categories. All counts were performed by the same observer and in a blinded fashion, and inaccuracies should have thereby been averaged out due to the sample size. Moreover, no entirely healthy cartilage was assessed, rather only the cartilage from patients who had undergone total knee arthroplasty for OA.
The "healthy" reference tissue was also derived from osteoarthritic cartilage. OA is, however, a condition that affects the entire joint.
Nonetheless, it is noteworthy that in many cases the results are clear and highly significant. Analysing entirely healthy cartilage as a baseline might have therefore yielded even stronger differences than those described in the present study. It also needs to be made clear that conclusions drawn from the immunohistochemical analyses were based on the signal intensity under the microscope and from the ELISA on the band strength. Both techniques depend on specific antibody binding to epitopes present in the tissue. Indeed, while in many cases a difference in signal or band intensity also signifies a change in protein content via a change in available epitopes for the antibody, it is also conceivable that epitopes get selectively destroyed while the core protein itself is still left in place. 49 It is similarly possible that the proteins get fragmented and lose their function, yet the epitopes remain in the tissue thus allowing antibody binding to still take place.